Re: MC PHONO PREAMP



Ian Bell wrote:

Patrick Turner wrote:

Ian Bell wrote:
Patrick Turner wrote:
Ian Bell wrote:
Patrick Turner wrote:
Ian Bell wrote:
snip
I am operating 6CG7 at 5mA where gm is just a shade over 2mA/V which
gives a calculated effective input noise resistance of 1250 ohms which
is about 0.6uV. The best 6CG7 I have measured so far gives just over 3uV.

So what you get theoretically is only pie in the sky and most days the
tubes are more noisy than the data and calcs suggest they ought to be.

Too bloody right.

gm does obviously have some bearing on the noise.

Pass. I guess one would need to measure tubes with large differences in
gm to be sure about that.

You can see what the gm is for a triode by looking at the typical Ia Ea
curves showing Ra for various Eg1 bias voltages.

Draw a vertical line through a wanted Ea for the tube and plot where Eg1
voltages intersect this line.

For say 6CG7/6SN7 at 10mA of Ia and Ea = 150V, gm is fairly constant
about a small Ia change of +/- 1mA/

But if you have Ia at 3mA, and Ia changes +/- 2mA, then gm changes a
lot.

The change in gm relates to a cube root of something squared.

With pentodes, the gm is more easily seen with a vertical load line, ie,
load that is a short circuit.


One has to measure the noise after an amp raises the signal at the anode
of a tube under test x 1,000 times then divide the noise figure by a
total of [1,000 x tube gain]. We assume the noise at the anode of the
tested tube is mainly all due to EINR. Its difficult to measure it any
other way.
Depends what you measure it with. My Ferrograph test set has a 1mV range
so a 40dB amp would make its full scale 10uV. Even a tube stage with
0.5uV EIN and a mere 20X gain will produce 10uV of noise at the output.
Hang on, if you had 0.1uV and 40dB of gain, ie, 100x, then you'd have
10uV after the gain.

But you don't have 0.1uV, even at the input. Say we have a perfect tube
with gm =2mA/V at the operating point and a gain of 20. Its input noise
will be 0.6uV so the noise you measure at its output will be 12uV.

OK
There is 1000uV per mV so the 10uV would be not a huge meter swing, only
1/100 of the linear scale.
Indeed, that's why I am suggesting a 40dB amp before it.

OK
But if your meter can tell you that you have 10uV then fine, but perhaps
its better to amplify the 0.1uV x 1,000 and get 0.1mV, and then you'd
see a swing of a meter needle of 10% of the scale on a 1mV range.

But it's not 0.1uV, its 12uV so you anly need 40dB of gain.

12uV x 100 = 1.2mV. If you have a 1mV or 3mV range then fine.

My mV meter I made long ago is only 10mV for the lowest range.


Ah, that explains why you need the extra gain.

The only time I figured I'd need to have a lower range for voltage
measurement is for distortion measuring.

So when I measure THD on 1kHz, I filter the THD out with a bridged T LC
notch filter and then amplify the THD with a bandpass amp based around
an opamp with gain = 10 and band from 1.6Khz to 10kHz with sharp roll
off below 1.6kHz to get some extra 1kHz attenuation, and obstruction to
all noise below 1kHz.

Therefore if I have 0.1mV of THD, it appears as 1mV on my meter, and if
the signal tested is 1Vrms, this means I have 0.01% THD, which is a
common typical amount of THD for an SE triode preamp with a µ-follower
topology.

If V tested = 10Vrms as from say a good measuring SS amp the THD is
0.001%, a fairly typical result.

But apart from THD measurements, having a range 0 to 1.0mV is a bit
useless, and very prone to noise.




When testing 1/2 12AX7 with shorted grid via a short lead, and if you
have 2uV of EIN, then expect 120uV at the anode if the AX7 gain is 60x.
In my case the following amp has gain of 1,500, so this becomes 180mV,
and a 1V scale is OK.
Yes, but my Ferrograph has a 1mV scale so I don't need anywhere near as
much extra gain.

Indeed...
The input noise of the amp used to lift the signal
to measure it
is maybe 3uV or less and so the noise at the amp input is the square
root of ( 120uV squared + 3uV squared ) and if you work this out then
the following amp noise makes virtually no difference One could have its
noise at 10uV and the 12AX7 noise would swamp the noise of the following
amp. My amp I use is a 2 stage tube type. One could use an opamp.

My noise measures are all with dc on the heaters.

Agreed

With ac on the heaters expect much more hum often due to leakage current
between cathode and heaters.
Yes, been there, done that, pulled my hair out and bought the T-shirt.

All my preamps always use dc heaters to avoid cathode hum.

A bad tube will have intermittent "sploshy" noise whose amplitude varies
between good and bloody terrible, and so you have to bin thse tubes
unless where you use it doesn't matter, maybe a second stage in a guitar
amp.

Selection for low noise is definitely to be recommended for the first stage.

That was one of the techie's job when he visited a studio or radio
station for regular maintenance, or when a tube became noisy.
In the old days the recording studios may have had quite a number of
12AX7 in consoles etc and a tech would have to swap out tubes to keep
the noise low at service time. Of course the EF86 became immensely
popular because you could get more gain than with a triode and then use
NFB to help control noise. RDH4 has a few typical ideas used in studios
and broadcast stations. One is to use a pentode with its anode taken to
0V and then use the screen as the anode with a lowish "anode" current.
Yes, I am well aware of that one. I even managed to get a hold of the
article the RDH4 refers to. The best noise I have got so far was from a
6AU6 pentode wired as a triode as per this article.

Maybe the shot noise is less because the metal area the electrons have
to land on is much smaller than the much larger area of the official
anode.

That's a very good point. The screen grid is clearly much smaller than
the anode. However, the papers I have read say shot noise is due to
changes in the rate of arrival of electrons which should be unchanged.

Probably right. Temperature needs to be kept low though or else the
secondary emission may make noise higher. So the Ia using g2 as anode
cannot be too high, which means gm will be on the low side as it varies
a lot with Ia.



Good screening, but I assume the EIN is still related to the 2.5/gm, and
gm of this sort of triode connected pentode is rather low with typical
types,
I have measured gm between 3 and 4mA/V for a 6AU6 in triode mode.

So it should be because that's the pentode gm.

When screen and anode are connected to a fixed B+, and a 10 ohm anode
current sensing R connected, the gm can easily be measured by biasing
the g1 for a given idle Ia and then altering the bias +/- with a signal
and measuring the Ia change. Because Ea and Eg2 doesn't change in the
test, triode and pentode gm should be the same.


although perhaps with 6EJ7 whichy is a high gm tube, or an E280F,
then maybe they are quieter than normal full triode connection. I recall
I tried the grounded anode idea but found magnetically induced hum to be
a bother with EF86.

I have not read any reports on the Net or in old books where anyone
successfully managed to get tube input noise lower than has been
mentioned as typical in RDH4.

Me neither. I have an article by Burgess from the Wireless Engineer
February 1945 which quotes the 2.5/gm derived by Harris and says it
'shows close agreement with measured values'. Now I think Harris and I
know for certain the Burgess is talking about radio receivers and for
them the lowest noise stage needs to be the rf amp or mixer.

Yes, indeed, especially when you are trying to sift out a low level
signal from all the RF noise picked up and the noise of the receiver
itself.

But with AM sets RF amps or mixer noise doesn't matter much for
broadcast stations.

Now maybe
2.5/gm applies at rf frequencies but it looks to me like it does not
apply for audio.

Why would it not apply?


Maybe it does but maybe at audio frequencies other mechanisms begin to
dominate. I don't know. What I do know is they did not do their testing
at audio frequencies so we cannot be certain the formula applies.

Unfortunately we don't have access to the references nominated in RDH4
about such issues around tube manufacture and R&D.
Many people have a copy of RDH4, but that's all, and they don't have the
library full of other material which delves deeper into tube behaviour.
Most has been chucked out of libraries long ago.

Page 783 of RDH4 states equivalent input noise 2.5/gm and that this is
an expression of shot effect noise based around the idea that such noise
is equivalent to an R between grid and cathode. It is of course an
imaginery R because grid input resistance is in megohms,
although grid input impedance is falling as F rises due to C.

A useful table appears on page 938 with input noise resistances given
for various tubes. The lowest EINR is for 1/2 a 6J6 which is 470 ohms.
And so if you used both triodes of a 6J6 you'd have EINR = 235 ohms and
about a third of the noise of a 1/2 of 12AX7.
And I will eat all my 12AX7s if you can find me a 6J6 with noise much
lower than a 12AX7. EC97 should also be good, but samples I have are
not.

I don't see where RDH4 says the noise depends on frequency, or that the
EINR was calculated based on measurements at RF.

The rules about noise are valid right down to dc once you have said that
the noise is due to a resistance.

So where you have bandwidth of dc to 20kHz, and say you have an R giving
2uV of noise, then if R is reduced by 10 then noise reduces by 1/3.16 or
about a third. And if the bandwidth is effectively reduced from 20khz to
50 as it is with an RIAA filter then noise is reduced the square root of
the amount of bandwidth reduction, ie, sq.rt ( 50 / 20,000 ) = 1 / 20,
or -26dB.




And wouldn't frequency vary things? as F rises, Rin drops, due to input
C.

Yes, perhaps, which would make things worse a low (audio) frequencies.

Depends on the resistance. The lower the R the less effect C has.

Tubes tend to have higher noise as F is lower, and this spoils their
effective use for phono inputs because of how RIAA eq means bass signals
off records are low and amp gain must be high at LF.

j-fets on the other hand have less LF noise and very little rumble or
flicker noise.

Now if the record industry had settled for 18" records with 45 RPM,
maybe things could have been different....

They ended up finally settling for a 5" silver disc with 1 and 0 used in
a sort of morse code........

Now the disc is doomed, and its all done by file handling on memory
chips. Not even a hard drive disc involved.

But I still enjoy a good vinyl.

Patrick Turner.



Indeed, Burgess references Harris' paper which is
entitled 'Fluctuations in space charge limited currents at MODERATELY
HIGH FREQUENCIES' (my capitals) which does no sound like audio to me.

Nobody bothers to really examine the issue of tube input noise very much
in these times of opamps and fets and so forth.

But it is something WE should examine as tube users.

Indeed. But only if we need to. I test tubes for noise to make sure my
customers are not upset by noise....

Many sellers and traders of tubes have no technical abilities and they
trade NOS tubes that are anything but NOS and are infact pulls that look
well in a box that looks new after having been carefully stored.

NOS tubes are often a complete scam.


But I found that afer several years of business I hade acquired a
collection of preamp tubes, mainly pulls from garbage gear, and I began
to wonder what was actually good amoung the pile. Some had to be good if
the garbage gear had had its tube replaced just before it was junked.
And I found some tubes were, low EIN, not very microphonic, nice bright
looking getter, hmm, maybe very close to NOS, so I went through the
piles and soon discovered what I could put in a "good" pile, ie, less
than 3uV for AX7, AU7, AT7 CG7, and what would be N, for noisy with the
average per section written on the tube, and M, for microphonic, ie,
well above a good sample. Some were quiet but M. Some are marked N,M,
meaning a bit noisy and microphonic but usable in a power amp second
stage for good while. Quite a few were just trash, more than 10uV and
micro, so out they went to the bin. Some had terrible hum.

Can you remember what proportion of the total had low noise and
microphony?/

Maybe 20%.


Interesting. I would have expected a lower figure but 20% is quite
encouraging.

One non technical guy had inherited an enormous number of 12AU7 and 6CG7
amoung cratefulls of stuff an old guy had left over from being a tech.

He offered them to me and I explained that old guys' collections of
tubes are usually crap and that I couldn't pay much.
OK, so we negotiated a deal and a price based on me paying a certain
agreed number of $$$ per tube if they tested OK. He would send them and
I would test them and any that failed my tests would be sent back with
noise and microphonic figures on the wrappers.
So a box arrived and I tested each and every one of about 150 tubes.
Took me quite some time, and at tech rates this added up to a
considerable sum. Most of the 6CG7 were in fact brand spanking new. But
only 20% of teh 12AU7 were OK, so a pile went back with my cheque for
the goodies.

The 6CG7 were made in Oz and I am content they sound as good as anything
by RCA or Seimans, but they don't fetch the big E-bay prices based
largely on the perceptual bull*** of the hi-fi glitterati.


I recently bought some 6CG7 off eBay and all the RCA ones tested better
than the others.

In the earlier era of 1930s to 1960s, Oz had its own protected
electronics industry so we could be independant if need be.

So we made our own range of tubes and they were mainly very good because
the industry was honest and the working girls were well paid in nice
conditions so they did good work.

Now all that's gone, and won't be again.

Now the US is hard at it making thousands of robot thinges for soldiers
to use for its security such as something the size of a model aeroplane
and crammed full of secret electronics is the go.

No need for us to have a local electronics industry for national
security.


LOL

The terrorist in the open field waiting to make a hit now has a much
decreasing chance of succeeding. Some little gizmo will fly high above
and unseen, and guide in a little surprize package.....

But of course we have seen superior technical force defeated during
history......


Oh, how true.

heers

Ian

Patrick Turner.
.

Quantcast